preload
Join the Directory | Contact Us
 
Bookmark and Share
Interview: Bill Fox, National Center for Patient Interactive Research 20 Most Popular Health Websites
Dec 06

Social Media Faceoff: Lipitor vs Crestor

healthcare marketing, research, social network marketing Add comments

Kru Research used the ListenLogic social media monitoring platform to monitor social media comments and measure sentiment for two cholesterol lowering drugs, Lipitor and Crestor, for one month (August 2009). An influence ranking of 5+ was used to reduce the amount of spam and irrelevant posts.

The questions we sought to answer were:

1)      What is the total number of comments for each of these brands?

2)      Which sites generate the most chatter (eg, Twitter, Blogs)?

3)      What is the overall sentiment score for each brand?

4)      Is there a difference in the topics or issues that are being commented on for each brand?

Share of Voice: Lipitor beats Crestor 2.3 to 1

After tracking 180 million websites and cleaning out the promotional chatter, we see that market class leader Lipitor had 695 total mentions and Crestor had 302 mentions. Lipitor has been on the market longer than Crestor, and is considered the best selling drug in the world, so it isn’t surprising that it has more activity on the social networks.

Little Difference in Sentiment-Mostly Neutral

When you look at all the conversations and comments about Lipitor and Crestor the vast majority, about 94%, is labeled “Neutral.” This is because most mentions picked up by listening platforms has to do with general corporate news, lawsuits, and investor related articles that mention the blockbuster drugs as part of their description of Pfizer and AstraZeneca. When you look only at health related comments, the picture changes.

Lipitor’s positive sentiment was approximately 9% versus Crestor’s 1%. Negative sentiment was 27% for Lipitor and 18% for Crestor.

Another way to look at sentiment is the ratio of positive to negative comments. Using this approach Lipitor scores .33 to .05 positive to negative sentiment.

Comment Cloud Analysis

The cloud tags shown below reflect that the brand names are being used in a variety of discussions that go far beyond the management of cholesterol.

cloud_lipitorcloud_crestor

Websites with Most Activity

Using a 5+ influence rating corrects for a lot of irrelevant chatter that happens on Twitter. However, both products unfortunately have most of their social comments happening on pharma gossip site CafePharma. Major sites that these chole

Of the 695 comments for Lipitor:

  • 46 (7%) came from CafePharma.com
  • 36 (5%) came from wikio.com
  • 22 (3%) came from medhelp.org
  • 18 (3%) came from Boards.webmd.com
  • 13 (2%) came from weightwatchers.com
  • 11 (2%) came from Twitter.com
  • 10 (1%) came from blogspot.com
  • 9 (1%) came from diabetesdaily.com
  • other

Of the 302 comments related to Crestor:

  • 17 (6%) came from CafePharma.com
  • 12 (4%) came from medhelp.org
  • 11 (4%) came from technologyquestions.com
  • 10 (3%) came from wikio.com
  • 7 (2%) came from Boards.webmd.com
  • 6 (2%) came from blogspot.com
  • 5 (2%) came from weightwatchers.com
  • 5 (2%) came from diabetesforums.com
  • 4 (1%) came from Twitter.com
  • Interesting analysis. I like the idea around these "faceoffs." One concern...I was intrigued by the different sources of mentions and where they came from, so I looked up a few. The one that I found the most odd was technologyquestions.com. I know that you mentioned that Listen Logic used "an influence ranking of 5+ was used to reduce the amount of spam and irrelevant posts."

    So, I was curious why people were talking about Crestor at technologyquestions.com. I was hoping to find some interesting debate on some new piece of technology Crestor or AstraZeneca was using only to find quite the opposite. Every listing of Crestor on technologyquestions.com is spam. Here's a link for a quick search for Crestor on that site: http://bit.ly/5eNCBD. Unfortunately, the same can be said for almost all of the Wikio.com mentions for both Crestor (http://bit.ly/4uZlpe) and Lipitor (http://bit.ly/6IHLeT).

    I don't think I need to mention that CafePharma is not at all a healthcare site and instead is populated with posts almost entirely done by former and current sales reps, which doesn't give any real sense about the medical discussion happening around either of the brands.

    I wonder if another monitoring tool might give some cleaner results or if the "influence ranking" isn't set high enough.
  • kruresearch
    Jonathan, thanks for taking the time to dig in and you bring up some really important points I think when it comes to "listening." And I'd love to use a parallel listening tool if they want to give me access to it! (the term "face-off" would then have a double meaning)

    The influence rating algorithm in ListenLogic, and in some other platforms I've seen, I think are independent of the content type. Meaning an author may have more influence by the number of followers, and number of their followers' followers, on Twitter. Or using in-bound link analysis etc. for blogs. But just because a source that is picked up has high influence doesn't mean it's a medically relevant source or patient-relevant comment.

    And in fact I find with big brands--and Lipitor is of course about the biggest--the vast majority of online chatter is about the company, the stock, the revenue of the drug and NOT medically related or by or for patients.

    So that gets to filtering and screening of course. The process I used isn't a good real world application in that I'm doing quick screens on a single month. ListenLogic has both it's own filter and analysis of spam, but then there is a "human analyst" component that can go in and review and clean it up further for what gets through. The system "learns" over time as it gets smarter from the decisions of the human analyst. As you know, some listening platforms just use human analysts which generally are accurate but take time and can be costly. Other platforms use just machine logic, which is fast, cheap and inaccurate. ListenLogic uses a combination of the methods.

    So back to the issue of spam and non-medical brand mentions. In this August study I attempted to cut the spam but clearly failed to do a good enough job. I could have cranked up the influence meter higher which might have cut out a bit more of it but should have also spent more time double checking the output.

    Also, thus far, in these face-offs I've left the non-medical mentions in thinking that the sentiment of a brand should include what positives/negatives that come from other sources. BUT, I've learned that the vast majority of of chatter is non-medical and neutral - which isn't very helpful in a real-world application. Measuring medical content only would of course greatly reduce total volume of mentions and increase accuracy related to patient.

    Lastly, and maybe another post or article for the both of us, I'm increasingly thinking that sentiment analysis for pharma brands really isn't that important at all. Certainly not the way it is for consumer goods brands. What's more relevant are identifying the themes that emerge from the content, the discussions, and (of course) identifying influencers and patients who you want to build a relationship with.

    Thanks again for weighing in.
    - Kevin
blog comments powered by Disqus

© 2010 Kru Research, LLC. All Rights Reserved.